Netflix Founder Reed Hastings on 'Skills That Matter in the AI Era'—Reading the Future of Education, Work, and Investment
Reed Hastings, the co-founder who led Netflix for 25 years and defined the streaming era. Since stepping down as CEO, he has stood at the forefront of AI safety as a board member of Anthropic, and has viewed the tectonic shifts in technology across the board through his experience as a board member at Microsoft, Meta, and Bloomberg. He earned a master's degree in AI from Stanford University in 1988, making him a rare individual who has experienced both the 'previous boom' and the 'current boom' of AI.
Based on his remarks on the podcast 'Possible,' hosted by Reid Hoffman (co-founder of LinkedIn), we organize his thoughts on entertainment, education, employment, and national strategy in the AI era.
1. After stepping down as CEO—'I was surprisingly fine'
1-1. Closing the curtain on 25 years
Hastings handed over the CEO seat to Ted Sarandos and Greg Peters in January 2023. He says his schedule vanished the day after he stepped down, and he spent February and March skiing and snowboarding.
He reflects that the biggest surprise was that 'the feeling of wanting to return hardly ever arose.' This was because he felt he had accomplished everything—the global expansion, the life of repeating overseas business trips almost every week. Of course, he has an attachment to the people, but he states that he 'maintains it as a personal connection,' frankly speaking that he had no lingering regrets about management.
1-2. Perspectives gained from diverse board experience
Hastings' board history is diverse. He has served on the boards of Microsoft since 2005, then Facebook (now Meta), and currently serves on the boards of Bloomberg and Anthropic.
He says his appointment to the Microsoft board was a product of chance. They reached out to big names in the tech industry, but were turned down one after another due to conflicts of interest, and the turn eventually came to the 'person in charge of a domestic DVD rental service' at the end of the list. At the time, Netflix was 'one-millionth' the size of Microsoft, but he hit it off with Steve Ballmer and Bill Gates during their meeting and joined the board. He says he learned a lot from Microsoft's stance of being able to invest in projects looking 10 years ahead.
He joined the Facebook board thinking that social media would transform Netflix, but he concludes that 'social had almost nothing to do with movies or TV shows.' Regarding his current role at Anthropic, he describes it as 'very exciting and tense,' emphasizing the value of having a seat at the forefront of AI.
2. Satya Nadella's 'one bold decision'
Hastings analyzes why Microsoft's corporate value has grown 10 to 15 times. He points out that while Office was stable, Windows was flat, and Bing search was struggling to grow, the standout success was Azure, and the biggest factor that grew Azure was AI workloads.
And what made that possible was the single decision of Satya Nadella's 'investment in OpenAI in 2018.' 'Satya made one incredibly bold and insightful decision,' Hastings evaluates. This investment brought legitimate workloads to Azure and became the driving force for Microsoft to catch up in a market where AWS had been ahead with Amazon's own primary usage and early adopters (like Netflix).
3. The 'missing question' in the AI debate—whether it's 18 months or 6 years doesn't matter
3-1. 'What to do after it arrives' rather than 'when it will arrive'
Hastings dismisses the debate over whether AGI will arrive in 18 months or 6 years as not very meaningful. 'It is coming fast. That is enough. How do we want to design society 10 or 20 years from now?'—he argues that this is the question that should be asked.
Citing the legal world as an example, he points out that arguments at the Supreme Court are conducted in almost the same format as they were 100 years ago. He states that even 20 years from now, when AI is fully integrated, the creation of briefs will be streamlined by AI, but the arguments themselves may not change essentially. On the other hand, there are fields that will change significantly. What is important is to understand that the impact of AI varies greatly by industry.
3-2. The field least likely to be affected: Entertainment
Surprisingly, the field that Hastings sees as having 'the least impact from AI' is entertainment, a field he has been involved in for many years. Using the intuitive analogy that 'no one wants to watch a game where robots play basketball,' he argues that experiences rooted in human emotion are difficult to replace with AI.
The appeal of March Madness (the NCAA basketball tournament) lies in the conflict between humans, just as the joy of giving real flowers does not change with AI. His basic assessment is that 'things that belong to the realm of emotion are relatively less affected by AI.'
3-3. The field most likely to be affected: Lawyers
Hastings cites lawyers as the profession most likely to be affected. This is because the work is highly dependent on language and follows certain patterns. However, he balances this by noting that if latent demand for legal services (such as the lack of legal services for low-income groups) becomes apparent, employment could potentially be maintained through an 'elastic response'.
4. Lessons from radiologists—'AI doomsday theories' are often wrong
A case study Hastings repeatedly refers to is radiology. Image processing is a field where AI outperforms humans, and four years ago, it was predicted to be the 'first profession to be decimated'.
However, the reality was the opposite. The price of an MRI scan dropped to $300 out-of-pocket, and the number of examinations skyrocketed. While a system was established where AI reads the images and radiologists approve them, there is currently a labor shortage due to the expansion of demand for examinations—with demand for about 40,000 radiologists against a supply of 35,000, and wages remain at a high level.
'We love Armageddon movies. Religions have them too. That's why we are drawn to scenarios where AI destroys something. But that hasn't happened in radiology,' says Hastings, warning against an excessive inclination toward catastrophic predictions.
5. AI Safety—Thinking in three categories
As a board member of Anthropic, Hastings is deeply involved in AI safety and organizes the risks into three categories.
First, the 'Skynet type'—a scenario where AI dominates humans. This will not happen in the short term, but as AI comes to handle more of our lives, there is a danger of 'sliding into' it. Like a large-scale nuclear war, he argues that preventive measures are essential because even if the probability of occurrence is low, recovery is extremely difficult.
Second, 'malicious use'—cases where terrorists or hostile nations use AI as a weapon. Specific risks include virus design through a combination with synthetic biology, or cyberattacks using AI (such as discovering vulnerabilities through open-source code analysis).
Third, industry-wide technical preventive measures—currently, each company is taking measures voluntarily, but Hastings sees the possibility that regulations will be needed for all sufficiently powerful AI systems in the future. However, he states that since the second category is not enough to 'destroy all of humanity at once,' a phased approach of building defense systems after accidents occur is also realistic.
6. Entertainment and AI—The illusion and reality of 'democratization'
6-1. The 'democratization of film' repeated for 50 years
Hastings points out that the democratization of filmmaking has been talked about for 50 years. The digital filming revolution of the 1990s also promised lower costs, but in reality, budgets increased and special effects just became more sophisticated. 'The constraint was never actually the cost of film. Student films were made in large numbers then and are still made today, but they don't break out.'
6-2. Lessons from K-pop Demon Hunters
Hastings reveals that even Netflix's recent big hit 'K-pop Demon Hunters' was the company's 28th animated work. He says that replicating success is extremely difficult and is 'like lightning striking the same place twice.' AI lowers the costs of 'mechanical and industrial parts' such as VFX and stadium crowd shots, but the ability to create the core of a story—character arcs, conflict, resolution—is a different dimension of the problem.
6-3. Will AI accelerate 'excessive sequelization'?
To the concern that if data-driven decision-making is enhanced by AI, reliance on past success patterns will strengthen and original works will decrease, Hastings is negative. Looking at the volume of new content Netflix produces, he states that 'excessive sequelization is not actually happening.' People seek both 'familiarity' and 'freshness,' and the balance of that tension is the essence of entertainment.
7. Education—The 'STEM era' might be over
7-1. What should be taught?
The theme Hastings emphasized most strongly is education. He divides the question into two parts.'What to teach' and 'how to teach'.
Regarding 'what to teach,' Hastings' argument is clear: 'If I had a 3-year-old child now, I would invest everything in emotional skills.' For the past 25 years, society has promoted STEM (Science, Technology, Engineering, and Mathematics). 'Learn to code' has been the mantra for years, but that premise is wavering due to the evolution of AI coding tools.
Hastings reflects that STEM has been 'overdone,' and he predicts that from the current state where Stanford University is entirely focused on STEM, there will be a return to the humanities—history, literature, brain physiology, and the understanding of human relationships.
7-2. Alpha School: The 'Tesla Roadster' of Education
As an advanced example of AI education, Hastings highly rates Alpha School. The school's philosophy is that 'children should want to go to school more than they want to be on break.' They solidify the basics with two hours of software learning every day, and the remaining time is devoted to autonomous activities such as sports, watching TED Talks, and discussions.
Hastings compares Alpha School to the 'Tesla Roadster of AI education.' Just as the original Tesla Roadster was a niche, high-priced sports car, it was the entity that made electric vehicles 'cool.' Similarly, he sees Alpha School as a pioneer proving the potential of AI education, and expects that a low-cost version equivalent to the 'Model 3' will eventually emerge.
7-3. Education Revolution in Developing Countries: 'One Laptop Per Child' Was Just 20 Years Too Early
In low-income countries, it is not uncommon for education budgets to be $300 per student per year, with class sizes of 50 to 70 students. Hastings argues that the combination of Starlink for every school, a tablet for every student, and excellent AI software can significantly narrow the education gap with developed countries.
To critics who cite the failure of 'One Laptop Per Child,' he counters that it is the same structure as the failure of expert systems in the 1980s. 'Just because it failed before doesn't mean it won't work this time. It was just 20 years too early.' He states that the combination of a $50 smartphone, Starlink, and solar panels is sufficiently scalable.
8. Employment and Wages: Determined by 'Supply and Demand,' Not 'Value'
8-1. Robot Plumbers Will Still Be Under 1% in 20 Years
Hastings calmly evaluates the impact of AI on manual labor. Citing autonomous driving, he points out that 20 years after the 2007 DARPA Challenge, autonomous driving still accounts for less than 1% of global mileage. He offers a similar outlook for robotic plumbing, stating it will be 'at most 1% in 20 years.' His view is that blue-collar technical jobs will remain promising careers for the next 20 years.
8-2. Wages Are Determined by 'Supply and Demand,' Not 'Value'
In response to a question from Reid Hoffman, Hastings introduces an important distinction: 'Teaching is a very valuable job, but it does not command high pay. Wages are determined not by value, but by the scarcity of supply and demand.'
Human wages will fall for administrative and clerical tasks that AI excels at, while emotional and interpersonal tasks that AI struggles with will continue to maintain high wages. This framework serves as a practical guide for thinking about career choices and career strategy.
9. Geopolitics: The Harsh Reality of 'Middle-Tier Nations'
9-1. AI Will Widen Income Inequality
Hastings states bluntly that AI will widen both income inequality within the United States and the gap between the U.S. and other countries. It is a cold-eyed view that 'AI will be dominated by China and the U.S. There is no answer for middle-tier nations.'
In an analogy to the Industrial Revolution, he cited the history of how Britain intentionally blocked the industrialization of Argentina and India, saying, 'Estonia promoting good government technology with digital IDs is better than doing nothing, but I don't know if it will be a real answer for middle-tier nations.'
9-2. Is an 'Age of Abundance' Coming?
On the other hand, I am optimistic in the long term. Technological revolutions catalyzed by AI—such as nuclear fusion energy, the dramatic reduction in housing costs through robotic construction, and 3D printing—have the potential to lower costs across industries and usher in an 'era of abundance.' However, he also notes the failure of early optimism regarding nuclear energy, which predicted that 'electricity would be too cheap to meter,' and does not forget to remain cautious.
Hastings concludes that the most important factor is the political mechanism for sharing those benefits. 'If everything goes well, it is because AI has unleashed human prosperity and we have found political mechanisms to share the benefits among income groups within countries and between nations.'